diff --git a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/Analyzer.scala b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/Analyzer.scala index dfa4215f2efe54b7526012e867a40155b1cb4e96..c239e83271615fac5b7b6519d5f53f3f4b23b7ef 100644 --- a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/Analyzer.scala +++ b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/Analyzer.scala @@ -561,6 +561,21 @@ class Analyzer( /** Extracts a [[Generator]] expression and any names assigned by aliases to their output. */ private object AliasedGenerator { def unapply(e: Expression): Option[(Generator, Seq[String])] = e match { + case Alias(g: Generator, name) + if g.elementTypes.size > 1 && java.util.regex.Pattern.matches("_c[0-9]+", name) => { + // Assume the default name given by parser is "_c[0-9]+", + // TODO in long term, move the naming logic from Parser to Analyzer. + // In projection, Parser gave default name for TGF as does for normal UDF, + // but the TGF probably have multiple output columns/names. + // e.g. SELECT explode(map(key, value)) FROM src; + // Let's simply ignore the default given name for this case. + Some((g, Nil)) + } + case Alias(g: Generator, name) if g.elementTypes.size > 1 => + // If not given the default names, and the TGF with multiple output columns + failAnalysis( + s"""Expect multiple names given for ${g.getClass.getName}, + |but only single name '${name}' specified""".stripMargin) case Alias(g: Generator, name) => Some((g, name :: Nil)) case MultiAlias(g: Generator, names) => Some(g, names) case _ => None diff --git a/sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/HiveQuerySuite.scala b/sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/HiveQuerySuite.scala index 089a57e25c08d21cdbae099ba44179e6a3ba235a..e7aec0b188c666f0846117d582796ff45e07648f 100644 --- a/sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/HiveQuerySuite.scala +++ b/sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/HiveQuerySuite.scala @@ -111,13 +111,13 @@ class HiveQuerySuite extends HiveComparisonTest with BeforeAndAfter { | SELECT key FROM gen_tmp ORDER BY key ASC; """.stripMargin) - test("multiple generator in projection") { + test("multiple generators in projection") { intercept[AnalysisException] { - sql("SELECT explode(map(key, value)), key FROM src").collect() + sql("SELECT explode(array(key, key)), explode(array(key, key)) FROM src").collect() } intercept[AnalysisException] { - sql("SELECT explode(map(key, value)) as k1, k2, key FROM src").collect() + sql("SELECT explode(array(key, key)) as k1, explode(array(key, key)) FROM src").collect() } } diff --git a/sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/SQLQuerySuite.scala b/sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/SQLQuerySuite.scala index e60d00e63574d612c02ebf5d8c937e1cb7aacd25..fbbf6ba5947dcfb9fb41120ae671fc4a22f80a33 100644 --- a/sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/SQLQuerySuite.scala +++ b/sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/SQLQuerySuite.scala @@ -548,13 +548,36 @@ class SQLQuerySuite extends QueryTest { dropTempTable("data") } - test("resolve udtf with single alias") { + test("resolve udtf in projection #1") { val rdd = sparkContext.makeRDD((1 to 5).map(i => s"""{"a":[$i, ${i + 1}]}""")) read.json(rdd).registerTempTable("data") val df = sql("SELECT explode(a) AS val FROM data") val col = df("val") } + test("resolve udtf in projection #2") { + val rdd = sparkContext.makeRDD((1 to 2).map(i => s"""{"a":[$i, ${i + 1}]}""")) + jsonRDD(rdd).registerTempTable("data") + checkAnswer(sql("SELECT explode(map(1, 1)) FROM data LIMIT 1"), Row(1, 1) :: Nil) + checkAnswer(sql("SELECT explode(map(1, 1)) as (k1, k2) FROM data LIMIT 1"), Row(1, 1) :: Nil) + intercept[AnalysisException] { + sql("SELECT explode(map(1, 1)) as k1 FROM data LIMIT 1") + } + + intercept[AnalysisException] { + sql("SELECT explode(map(1, 1)) as (k1, k2, k3) FROM data LIMIT 1") + } + } + + // TGF with non-TGF in project is allowed in Spark SQL, but not in Hive + test("TGF with non-TGF in projection") { + val rdd = sparkContext.makeRDD( """{"a": "1", "b":"1"}""" :: Nil) + jsonRDD(rdd).registerTempTable("data") + checkAnswer( + sql("SELECT explode(map(a, b)) as (k1, k2), a, b FROM data"), + Row("1", "1", "1", "1") :: Nil) + } + test("logical.Project should not be resolved if it contains aggregates or generators") { // This test is used to test the fix of SPARK-5875. // The original issue was that Project's resolved will be true when it contains